A group of researchers from the University of Pennsylvania recently published an interesting study in which they analyzed roughly 10 billion tweets to better understand how people are talking about cardiovascular disease on social media.

Twitter has had active conversations around health for years, including between clinicians and patients on health issues. Interest in mining that data is not new. A few years ago, we covered some interesting studies looking at prediction of adverse effects of new drugs based on social media posts. As far as big data goes, the flow from this particular firehouse is particularly brisk.

In this study, researchers from UPenn looked at a sample of about 8% of all tweets sent over a ~5 year period – that amounted to around 10 billion tweets. They searched for tweets in English that included keywords related to heart disease and associated risk factors including heart attack, hypertension, cardiac arrest, hypertension, and diabetes. That left more than 500,000 tweets. They then randomly selected 2,500 tweets (500 associated with each term) and manually coded them to describe content.

Around 80% of all of tweets related to cardiovascular disease were focused on diabetes and heart attack. Not surprisingly, 42% of tweets were focused on risk factors for heart disease. Interestingly, 35% of tweets included first person accounts related to heart disease but only 3% identified the user as actually having a heart disease.

While we frequently talk about how many of our patients are turning to social media to learn about health conditions, this study uniquely quantifies that conversation. I was particularly struck by the discrepancy between the tweets including first person accounts vs. self-identifying as having a heart disease. That seems to suggest that even more so than our patients, their families and friends may be the ones turning to social media for more information.

This study could also provide some helpful insights as clinicians and health institutions try to figure out how to engage with their patients through social media. A useful next step would be defining who is engaging with health messaging on Twitter and the kind of messaging that maximizes that engagement. For example, prior studies have suggested that teens don’t really trust health information on Twitter.

Reference: Sinnenberg et al. JAMA Cardiology. 2016. PubMed